Has anything changed in the way the TCS community publishes papers? Comparing recent STOC 2012 and FOCS 2012 to their counterparts STOC 2003 and FOCS 2004 one thing that has changed is that today we do not allow submitting printed copies by mail anymore. All submissions are done electronically.
But a much more radical change is happening in the last few years. More and more researchers actively prefer to have their publications posted on-line in a manner that allows free access to everyone. This change is reflected in the following sentences found in these recent CPFs:
“Authors are encouraged to post full versions of their submissions in a freely accessible on-line repository…”
“We expect that authors of accepted papers will make full versions of their papers, with proofs, available by the camera-ready deadline”
This is a bold step toward a world where academic publications will be freely available to everyone. There are several simple ways we can reach this vision together. First we should all encourage and expect from ourselves, our co-authors and our colleagues to follow these guidelines. Second, we should try to convince all major TCS conferences to adopt similar policies.
Why not go a step further? The next stage after encouraging and expecting is having some mechanism to enforce it. Our suggestion is simple: if the authors of an accepted paper fail to post an on-line version then the PC chair can post on-line the version that was submitted for review. Here is an example of potential CFP text:
“Authors should post full versions, with proofs, of their submissions in a freely accessible on-line repository…”
“Accepted papers that do not have on-line full versions by the time the conference ends will have their submitted version posted on-line by the PC chair.”
Note that we suggest extending the on-line posting deadline until after the conference ends, that is after the talk is given and the proceedings with the 10 pages “conference version” is printed (into paper/USB flash key/Internet). This may solve some patent timing issues and also allows authors the option to keep the community temporarily in suspense and provide incentives for physically attending the conference.
A quick check of FOCS 2011 accepted papers (starting with the list compiled by Shiva Kintali) shows that less than 10% of the papers do not have an on-line version today. So we expect that enforcement by the PC chair will be minimal (probably zero).
Should we as a community take such a step? What could the next steps be?
Ittai Abraham, Alex Andoni, and Kunal Talwar
Windows On Theory 
Great suggestion!
If only 10% of papers do NOT have an online version, why enforce anything ?
I don’t think forcing the issue is a good idea. Authors may want to keep their paper private for a variety of good reasons. Our conference system is capricious. For instance, one paper may get a preliminary result but has the seed idea which is very good. Some one else may pick up that idea and obtain a much stronger result. This is not a problem if the first paper is accepted but could be an issue if it is not accepted. Unlike archival based fora such as journals, conferences are not obliged to publish the seed idea papers. It is not always possible to know ahead of time whether one should or should not make a paper public. Reviews are one form of getting feedback before making such a decision. The fact that only 10% of FOCS 2011 accepts are not available is not good evidence because there is a selection bias there. One could say that the authors of strong papers put them up because they are confident that the result is complete and has a good probability of being accepted. What one should look at is the percentage of papers that were submitted but were not available on the arxiv.
Chandra,
I think you’re confusing the proposal in the post with the more radical proposal I raised at the STOC 2011 business meeting (which was that papers should be publicly available even before submission). They are proposing only that *accepted* papers will be forced to be posted on-line. In addition to ensuring public availability, their proposal will help eliminate the too-common problem that authors claim theorems in their conference papers but never get around to making the full proofs available. (A question I have about the 10% figure is whether the versions that appear online are just the 10-page conference papers, or full papers with proofs. If it’s largely the former, we still have a problem to solve.)
At a later point in the discussion, I may try to explain the rationale between my stronger proposal, but for now I don’t want to distract from the new proposal (which I support).
Salil
geomblog: you are right that the case of regulation is not obvious given the self-enforcement rate. I think it may be worth discussing though. Is ten percent low enough? The cost of regulation in this case is extremely small so the ten percent benefit to the community of making the works open access may outweigh the cost.
There is also the issue that Salil mentions of the truncated proofs-omitted-for-lack-of-space papers for which longer versions are hard to find. In cases when the full proofs exist in the submitted versions, this regulation may encourage authors to put something closer to full proofs when preparing the arxiv version. Of course authors may choose to just put the final conference version on the arxiv. Hopefully in cases when there is a fuller version under submission to a journal (or say being privately circulated amongs friends), such a version may appear on the arxiv.
There is also a hope that the discussion of “regulation” may by itself lead to more “self-regulation”. 🙂
Chandra: I agree that having all submissions being made public is a much more significant change, that will have affect in non-trivial ways the behaviour and the dynamics of the community. Predicting these affects and judging the desirability of those certainly requires more thought.
I may be wrong but I do not foresee any unintended consequences/behavioural changes of this mild proposal of putting accepted papers online after the conference. They are afterall already on acm/ieee websites soon after.
I think this is a great idea, and like Kunal I can’t really see the downside. Changing the number of available papers from 90% to 100% is not a negligible improvement.
Of course one can always criticize the proposal for being too mild, and you can think of stronger proposals, such as Salil’s, or a making a stronger push by our community to ACM/IEEE to make all our conference proceedings open access. (Including past conferences, after all even the “big bad Elsevier” opens up access to papers after X years from publication, but ACM/IEEE keep proceedings behind a paywall forever.) But this proposal is certainly a step in the right direction.
An even less radical step is for to make a public statement, perhaps by a symbolic vote in some business meeting, that as a community we consider it to be of poor form for an author not to post online a paper that has appeared in a conference or journal. I think that we are at a point where posting your papers online is no longer merely a nice thing to do, but part of our duties as scientists.
another point. maybe this was a slip, but hopefully your policy will require posting of the “final versions” of the papers, rather than the “submitted versions” ?
Secondly, this brings up an interesting conundrum. I prepare the submitted version of my paper, which is 10 pages of heavily compressed claims and an appendix that may or may not have fleshed out proofs. The paper gets in, and I have about a month to prepare a “final version” that is also 10 pages without any appendices because of antiquated dead-tree ideas of what a conference publication length should be, in some god-awful format that ACM insists I use. Then I ALSO have to prepare a full version with complete proofs (say in arXiv format) or else my crappy submitted version gets put online.
Which is fine, except that I really was trying to merge this paper with another one and submit to the Journal of Magnificient Results, so I need a FOURTH version of that, instead of merely posting the journal submission on the arxiv, because I’m picky about any content with my name on it that appears on the web.
I could go on, but my point is two-fold. If 90% of the papers are already online, then we’re dealing with edge cases, and for those edge cases, there are plausible reasons why I might not yet have the paper online. I’m much more in favor of public shaming (or cultural pressure) to encourage people to post papers online.
The PC chair is not going to look at what exactly every author posted, so I think an enforcable and reasonable policy should be that you must post *a* version of your paper online. If you don’t have anything better, just use the camera ready version that will anyway appear on the digital library. This is essentially zero effort on the part of the author.
Yes, this won’t necessarily solve the missing proofs problem raised by Salil, but at least 100% of the papers will be available online.
To clarify. I suggest the following policy:
Authors *must* post *some* version of their paper in some freely available online repository by the first day of the conference. Otherwise the submission version gets posted by the chair. The posted version is *expected* to contain the full proofs for the theorems in the paper, but this is not enforced.
This policy will increase the number of freely available papers to 100%, and I suspect will also increase the number of papers with full proofs as well. I believe at least in some cases, the reason that a full version is never posted online is that there isn’t any particular date they must do so by. From personal experience, some computer scientists have a hard time doing any task that doesn’t come with deadline..
Ittai/Alex/Kunal: why don’t you propose your suggestion to Tim Roughgarden? (FOCS 2012 PC chair)
For authors who don’t like papers being incomplete, one nice feature about arxiv.org is that it’s easy to replace papers with newer versions.
Boaz: I like your suggestion!
Since the FOCS 2012 CFP is already out, it may be difficult to change that now. I think Tim is aware of this discussion, and I would love to have him comment publicly on his viewpoint. Our suggestion is definitely aimed at any future PC chair of any major TCS conference.
BTW: Many of the specialized conferences (SODA/CRYPTO/EC/SOCG/PODC/etc) don’t even “encourage and expect” on-line posting (let alone enforce it). CCC is the only exception I am aware of! I wonder how we can encourage more conferences to support on-line posting.
I just noticed that STOC 2012, unlike FOCS and STOC 11, has a 20 page limit for the final version (though still in two column format). If more and more conferences do this, one would expect many more of the final versions to be self contained.
As geomblog suggests above, in most cases, this would mean one fewer version to prepare. Ideally this should make it easier for people to post arxiv versions.
SODA has had a 20-page limit for a couple of years, with an unintended consequence: Authors who try to publish their SODA papers in a (more) properly refereed journal are having their papers rejected because they declined to omit details from the SODA version.
The easiest way way to enforce the posting of every FOCS and STOC papers to public repositories would be to host the official FOCS and STOC proceedings on public repositories, instead of hiding them behind the ACM/IEEE paywalls. Why require an extra step by every individual author?